NEDOQwen: An Auditable Low-Cost Diagnosis-and-Repair Study for a Turkish-Centric 0.824B Language Model
A version-aware, low-cost workflow for diagnosing external evaluation failures, auditing answer-label bias, and reporting partial model repair.
Ethosoft Research
Explore our preprints, technical reports, controlled experiments, and reproducible artificial intelligence research.
Latest publication
Each page includes verified metadata, a complete abstract, citation details, and links to research artifacts.
A version-aware, low-cost workflow for diagnosing external evaluation failures, auditing answer-label bias, and reporting partial model repair.
A controlled study of the gap between learning a sentence-level future representation and using that signal to change language-model generation behavior.